How To Split A Bam File By Chromosome

Question

[GPR]

Hello,

I am having a hard time opening a very large bedgraph. I have been suggested to split my bam file by chromosome with chrom-bed.py but it didn't work.

Is there any other alternative?

Thanks,
GP

[Alex Paciorkowski]

How are you having a hard time opening the file? What happens? How did your script to split the bam file "not work"? These details may help people answer your question.

[GPR]

I am pretty much having trouble with the indexing and header of my bam file. I have indexed in various ways, but keep having trouble. Have tried chrom-bed.py and samtools view chr1 > chr1.bam. I have indexed with samtools and bamtools, I even tried sorting the file before indexing. Any suggestions?

[Madelaine Gogol]

bam != bedgraph, fyi.

[GPR]

I have indexed and when running samtools virew, I get the message fail to read the header from bam file. Any tips?

@ Madelaine: yes I have converted my bam files to bedgraph ones successfully. The problem is the size, that's why I want to split them by chromosome

[Michael]

Try samtools reheader to get the header back.

[Madelaine Gogol]

@genaro Oh, sorry.

You can also use picard CreateSequenceDictionary on the genome fasta and add that to the top of the sam file if the sam file is missing a header.

Answer 1

[Jorge Amigo]

in this other answer, Aaron Quinlan stated:

bamtools has a "split" command for exactly this purpose

I can only add that I've just tried it with this simple command

bamtools split -in file.bam -reference

and it works like a charm. the bam file gets split into different bam files, which are suffixed with .REF_xxx.bam by default, which is very convenient.

[mg]

Is it possible to specify output directory, and to mention which specify chromosome needs to be extracted (for eg., just chr5 and chr22)?

[ella]

Hi! Thank you Jorge Amigo for suggesting bamtools split. I successfully installed bamtools and am wondering, whether someone can answer if it is possible to not split the whole bam into all chromosomes/ contigs, but only extract those chromosomes/ contigs needed..?

[Jorge Amigo]

bamtools split is intended to split the whole bam file. What you are describing is definititely not what it was designed for, and it would be then much easier to extract the derised chromomoses/contigs directly with samtools.

Either by an individual samtools command:

samtools view -bo out.chr21.bam in.bam chr21

Either by nested samtools commands:

for chr in chr20 chr21 chr22; do
 samtools view -bo out.$chr.bam in.bam $chr
done

[russ.bainer]

This has the advantage of only generating a bamfile for the reference that you want.

Answer 2

[Michael]

Try samtools: samtools view -?

A region should be presented in one of the following formats: `chr1',`chr2:1,000' and `chr3:1000-2,000'. When a region is specified, the input alignment file must be an indexed BAM file.

something like samtools view in.bam chr1 > chr1.bam should work

[travcollier]

You're missing the -b flag (bam output). So something like:

​samtools view -b in.bam chr1 > in_chr1.bam

NOTE: The sequence dictionary (@SQ header lines) will still contain entries for everything. This can cause problems if the tools you're feeding those split bam files into use that header information. For example, picard CollectWgsMetrics will still assume that the bam is supposed to cover the whole genome, not just a single chrom. I'm certain lots of other tools will have similar problems.

[uribe.convers]

Is there a way to prevent each bam file to have entries for everything in the sequence dictionary (@SQ)? Or is there a way to recreate/parse/filter the header for each specific bam?

[Michael]

I would try the bamtools way (Jorge Amigo's answer) instead. It might be possible to parse and filter the entries going through text format, but it can also easily mess up everything.

[Chen Sun]

A solution would be converting bam to fastq file, then use the fastq to map to specific chromosome

[scchess]

This is silly as it's slow and the file wouldn't be indexed.

[ammar.husami]

need to specify the output format to be BAM with option -O BAM or with -b

Answer 3

[SHI Quan]

samtools view in.bam chr1 -b > out.bam

Use -b to output bam format

Answer 4

[Pierre Lindenbaum]

I wrote a java tool to split a BAM per chromosome see http://code.google.com/p/jvarkit/wiki/SplitBam

It also creates an empty BAM (filled with a pair of mock SAMRecords) for each chromosome in the Reference, if no SAMRecord was found for the chromosome.

Answer 5

[Shicheng Guo]

You can use the following pipeline to extract chrY reads from the raw bam files and with the header

samtools sort A.bam -o A.sort.bam
samtools index A.sort.bam
samtools view -H A.sort.bam > output.extraction.sam
samtools view A.sort.bam chrY >> output.extraction.sam
samtools view -hb output.extraction.sam > output.extraction.bam
samtools view  -H output.extraction.bam

output.extraction.bam is the bam file which extracted chrY reads.

[Jorge Amigo]

This isn't really an answer to the original question.

Plus if you want to extract chrY reads you only need a single command:

samtools view -b in.bam chrY > out.bam

[goodez]

This is true but still requires a sorted and indexed bam file. From the samtools manual:

Use of region specifications requires a coordinate-sorted and indexed input file (in BAM or CRAM format).

samtools sort A.bam -o A.sort.bam
samtools index A.sort.bam
samtools view -b A.sort.bam chrY > output.extraction.bam

[rohitsatyam102]

Using -b produces smaller bam than skipping it. Why so? For chrX, for example, I observed a bam of 2.1 GB with -b option but a 12 GB BAM. And they can be viewed with less or head command which is quite unusual.

Answer 6

[hewm2008]

This soft can help all of you

https://github.com/BGI-shenzhen/BamSplit

[Lhl]

well done. xiao-ming

Answer 7

[opplatek]

There is also a nice blog post about this by Sam Nicholls here.

tl;dr (extracted from the blog post)

samtools view -H in.bam | grep -P '^@SQ' | cut -f 2 -d ':' | cut -f 1 | while read contig; do
    samtools faidx reference.fa $contig > my_contig.fa
    java -jar picard.jar CreateSequenceDictionary R=my_contig.fa O=my_contig.dict
    java -jar picard.jar ReorderSam INPUT=in.bam OUTPUT=out-${contig}.bam REFERENCE=my_contig.fa S=true VERBOSITY=WARNING
    rm my_contig.fa my_contig.dict
done

This will go through the input whole bam file (in.bam) and make separate bams for all the contigs (out-${contig}.bam) in the reference fasta (reference.fa). The output bam files are compatible with most of the common bioinformatics software. Of course, you can skip the for loop and just use the contig name of your choice instead.

Answer 8

[etiennedanis]

samtools view -b in.bam chr1 > out.bam

did not work for me. But

samtools view -b in.bam 1  > out.bam

worked.

[Jorge Amigo]

Of course it depends on how your chromosomes are defined. If your bam file comes from a hg19 mapping you'll have to use "chr1", and if your bam file comes from a b37 mapping you'll have to use only "1".

[etiennedanis]

Good to know! I used mm10.

[Jorge Amigo]

You can always check for that information in advance by inspecting the bam file's header:

samtools view -H file.bam | perl -lne '/SN:(\S+)/ and print $1'

[etiennedanis]

Very nice! It does. I also found that

samtools idxstats file.bam

gives this info too. It is a good idea to check when using a new genome or a different version. Thank you!

[Dan]

How should I get more than 1 chromosome? samtools view -b in.bam [chr1, chr2] > out.bam ? Thanks

[Dan]

I found the correct code is samtools view -b in.bam chr1 chr2 > out.bam